On Sunday 29 January 2006 15:32, Tony Hoyle wrote: > Adrian von Bidder wrote:
> > Accuracy as in 'how close to 'real time' can you get' is just not an
> > issue. Reliability as in 'how much of the time is it available?' is.
> I'd have considered it the other way around... PCs have internal
> clocks, so can do without a reference clock for periods of time [...]
> OTOH if the source I sync to has clock drift & I end up 10 seconds out,
You're correct on both counts - BUT: accuracy to within a few tens of ms is
solved, it's not a problem. Servers are thrown out of the pool when they
reach something like 50ms offset (Ask - what value do you currently use?)
Since you should always use 3 or more servers if time synchronisation is
important to you the 'clock drifts out to 10 seconds' case just won't
happen in practice - unless you lose sync altogether, which is a
reliability problem.
Reliability, as in 'will the ntp server have the same IP over several
months' is imho the unsolved problem, especially in the pool where many
servers are operated by hobbyists and small entities.
cheers
-- vbi
--
Abwärts Tyrann, nach oben ein Knecht; Verleumder des Menschen,
Speichellecker des Herrn - voila des Glaubens Porträt.
-- Ludwig Feuerbach
pgpV9AI6gY10T.pgp
Description: PGP signature
_______________________________________________ timekeepers mailing list [email protected] https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
